All Times ET
Keywords: Gaussian process covariance, mixed effects models, spatial, temporal, Bayesian inference
We propose an efficient Bayesian approach for fitting zero-inflated negative binomial models with nearest neighbor Gaussian process for the varying temporal patterns. There has been much research of the urgent need model large spatial and temporal data due to the advances of spatiotemporal information and related computational technologies. Scalable spatial process models in the hierarchical Bayesian paradigm facilitate posterior sampling, we introduce a set of latent variables that are represented with nearest neighbor Gaussian processes, where the precision terms follow independent Polya-Gamma distributions in the row covariance and the column covariance from a Gaussian process. Our model accommodates multivariate and spatiotemporal data in which the existing approaches often fail due to computational challenges. Using simulation studies, we highlight the key features of the method and compare its performance to other methods. The approach is applied to model the the COVID-19 mortality data collected from various locations and time periods.